Article Preview
Top1. Introduction
Cloud computing is continuously evolving and becoming one of the most challenging paradigms of Information Technology with various business models to target the needs of users and enterprises (Buyya et al, 2008). Cloud providers are forming broad cloud ecosystem to meet the rising demands. In this paper, two types of cloud providers are discussed particularly the Service Providers (SPs) which ask for a service to be executed, and the Infrastructure Providers (IPs) which actually execute the service on their infrastructures.
With the rapid evolution of the cloud, together with the new emerging needs of customers, cloud environments are becoming very complex in terms of dimension and management. These systems are composed of various entities, such as stake-holders with different interests such as users or providers, Service Level Agreements (SLAs) contracts agreed between two parties, virtualised resources, to name a few. Thus cloud service models, whether being Platform as a Service (PaaS) or Infrastructure as a Service (IaaS), have huge amounts of information that needs to be collected, managed and evaluated for successful completion of the services. Therefore there is a need to define a consistent cloud-enabled data model which represents multiple entities and their interrelationships managing these efficiently. The virtualisation technology, in cloud environments, offers a manner to autonomously manage IT (cloud) entities allowing dynamic resource allocation that, based on accurate monitoring information, enables the appropriate enactment of cloud-related features like elasticity of services or high availability of resources.
Monitoring information can range from application-related metrics, such as web-based service response time, to infrastructure-related metrics like power consumption or resource capacity and utilisation. This information should then be assessed in order to provide the business-level parameters (BLPs) like trust, risk, ecological efficiency and cost metrics related to the service as a decision-making process.
Self-management of cloud systems should be governed by certain business-driven management policies which embrace the needs of both stakeholders: (1) users, who typically specify several constraints in SLAs, also known as Service-Level Objectives (SLOs); and (2) providers, who have their own Business-Level Objectives (BLOs) to fulfill such as saving costs and energy. This multi-purpose approach may lead to significant trade-offs, which must be solved by proper management policies. For instance, maximising the eco-efficiency of a cloud provider’s infrastructure is possible through the consolidation of several virtual machines in the same physical host. However, the performance of the services running in those virtualised resources may be diminished leading to the agreed SLAs being violated.
The assessment of high-level parameters, such as trust, risk, ecological efficiency and cost, are crucial inputs for these management policies driven by the business aspects in order to efficiently guide the operation of cloud providers in terms of the BLOs fulfillment.
To address the research challenges mentioned, this work contributes in several aspects:
- •
A generic cloud-enabled data model, which represents all involved entities in typical complex cloud environments and their interrelationships, presenting an efficient way to collect, aggregate and store monitoring information from cloud infrastructures.
- •
Several assessment tools that process monitoring information and are aimed to assist in the self-management of the cloud focusing on trust, risk, ecological efficiency and cost factors (Ferrer et al, 2012).
- •
Management entities for service (SP) and infrastructure (IP) cloud providers guided by decisions based on assessment tools.
- •
A policy-related approach driven by both customer needs and provider BLOs. This involves efficient information management and determining the most convenient management action(s) when IT-level events (expected or unexpected) take place.